Method for Detecting Defects, Electronic Device, and Computer Readable Medium

1. A defect detecting method comprising: acquiring an image of a test object; dividing the image of the test object evenly into a plurality of sub-images according to a size of a preset template image; determining, by a first model, whether each of the plurality of sub-images is similar to the template image, wherein if each of the plurality of sub-images is similar to the template image, determining that the test object has no defect, and if any of the plurality of the sub-images is not similar to the template image, determining, by a second model, whether at least one defect exists within the sub-image; and determining that the test object comprise at least one defect if at least one defect exists within the sub-image, wherein the second model is different from the first model.

2. The defect detecting method of claim 1 , wherein a process of determining whether each of the plurality of sub-images is similar the template image, comprises: matching each of the plurality of sub-images against the template image; obtaining a similarity value of each of the plurality of sub-images according to the first model; determining whether the similarity value is greater than a preset value, wherein determining that the sub-image is similar to the template image, if the similarity value is greater than or equal to the preset value, and determining that the sub-image is not similar to the template image, if the similarity value is not greater than the preset value.

3. The defect detecting method of claim 1 , wherein the first model is a similarity judgment model comprising a formula for calculating similarities between images, wherein the formula calculates a number of pixels which are same in two images, and then calculates the similarity value between the two images.

4. The defect detecting method of claim 1 , wherein the first model is a convolutional neural network model.

5. The defect detecting method of claim 1 , wherein the second model is a VGG model, or a ResNet model.

6. The defect detecting method of claim 1 , wherein a process of dividing the image of the test object into a plurality of sub-images comprises: searching for a boundary of the image of the test object; distinguishing a detection area and a non-detection area of the image; and dividing the detection area into the plurality of sub-images.

7. An electronic device, configured for detecting a defect on a surface of a detect object, comprising: at least one processor; at least one storage device storing one or more programs, when executed by the processor, the one or more programs cause the processor to: acquire an image of a test object; divide the image of the test object evenly into a plurality of sub-images according to a size of a preset template image; determine, by a first model, whether each of the plurality of sub-images is similar to the template image, wherein if each of the plurality of sub-images is similar to the template image, determine that the test object has no defect, and if any of the plurality of the sub-images is not similar to the template image, determine, by a second model, whether at least one defect exists within the sub-image; and determine that the test object comprise at least one defect if at least one defect exists within the sub-image, wherein the second model is different from the first model.

8. The electronic device of claim 7 , wherein a process of determining whether each of the sub-images is similar the template image, comprises: matching each of the plurality of sub-images against the template image; obtaining a similarity value of each of the plurality of sub-images according to the first model; determining whether the similarity value is greater than a preset value, wherein determining that the sub-image is similar to the template image, if the similarity value is greater than or equal to the preset value, and determining that the sub-image is not similar to the template image, if the similarity value is not greater than the preset value.

9. The electronic device of claim 7 , wherein the first model is a similarity judgment model comprising a formula for calculating image similarity, wherein the formula calculates a number of pixels which are same in two images, and then calculates the similarity value between the two images.

10. The electronic device of claim 7 , wherein the first model is a convolutional neural network model.

11. The electronic device of claim 7 , wherein the second model is a VGG model, or a ResNet model.

12. The electronic device of claim 7 , wherein a process of dividing the image of the test object into a plurality of sub-images comprises: searching for a boundary of the image of the test object; distinguishing a detection area and a non-detection area of the image; and dividing the detection area into the plurality of sub-images.

13. A non-transitory computer readable storage medium having stored thereon instructions that, when executed by at least one processor of a computing device, causes the processor to perform a defect detecting method, wherein the method comprises: acquiring an image of a test object; dividing the image of the test object evenly into a plurality of sub-images according to a size of a preset template image; determining, by a first model, whether each of the plurality of sub-images is similar to the template image, wherein if each of the plurality of sub-images is similar to the template image, determining that the test object has no defect, and if any of the plurality of the sub-images is not similar to the template image, determining, by a second model, whether at least one defect exists within the sub-image; and determining that the test object comprise at least one defect if at least one defect exists within the sub-image, wherein the second model is different from the first model.

14. The non-transitory computer readable storage medium of claim 13 , wherein a process of determining whether each of the sub-images is similar the corresponding template image, comprises: matching each of the plurality of sub-images against the template image; obtaining a similarity value of each of the plurality of sub-images according to the first model; determining whether the similarity value is greater than a preset value, wherein determining that the sub-image is similar to the template image, if the similarity value is greater than or equal to the preset value, and determining that the sub-image is not similar to the template image, if the similarity value is not greater than the preset value.

15. The non-transitory computer readable storage medium of claim 13 , wherein the first model is a similarity judgment model, the second model is a convolutional neural network model.

16. The non-transitory computer readable storage medium of claim 13 , wherein a process of dividing the image of the test object into a plurality of sub-images comprises: searching for a boundary of the image of the test object; distinguishing a detection area and a non-detection area of the image; and dividing the detection area into the plurality of sub-images.